#5 feat: add Copilot engine support via REVIEW_ENGINE toggle

Summary

Add a second LLM engine option (GitHub Copilot) alongside the existing Claude engine, toggled via the REVIEW_ENGINE repo variable.

Key changes

• scripts/engine.sh — New LLM abstraction layer with run_triage() and run_agentic() functions that dispatch to claude or copilot CLI based on $REVIEW_ENGINE
• scripts/review-one-pr.sh — Refactored to source engine.sh; engine-agnostic variable names (DELEGATION_ORGS, AI_DELEGATION_ENABLED, deep/audit tiers)
• All prompts — Engine-agnostic tier names, $ENGINE_LABEL footers, removed hardcoded model names
• Workflow — Conditional CLI install, REVIEW_ENGINE/COPILOT_GITHUB_TOKEN vars, DELEGATION_ORGS with CLAUDE_ORGS backward compat fallback
• Docs — Updated AGENT.md and README.md with both engine options

Model mapping

Tier	Claude	Copilot
Triage	Haiku 4.5	GPT-5-mini
Deep review	Sonnet 4.6	GPT-5.2
Security audit	Opus 4.6	GPT-5.4
Action/single	Sonnet/Opus	GPT-5.2/5.4

Usage

# Switch to Copilot engine:
gh variable set REVIEW_ENGINE --body copilot --repo don-petry/self
gh secret set COPILOT_GITHUB_TOKEN --repo don-petry/self

# Switch back to Claude (or leave unset — it's the default):
gh variable set REVIEW_ENGINE --body claude --repo don-petry/self

Backward compatibility

• Existing Claude setups work without changes (REVIEW_ENGINE defaults to claude)
• CLAUDE_ORGS is still respected as a fallback for DELEGATION_ORGS
• JSON field names in prompt output schemas (escalate_to_opus, sonnet_findings_*) kept as-is to avoid breaking the parsing logic